Preprocess Data for Schubert Winterreise Dataset

This notebook shows exemplarily how to compute input representations (HCQT) and pitch-class annotations for deep chroma estimation, at the example of the Schubert Winterreise Dataset (SWD).

© Johannes Zeitler (johannes.zeitler@fau.de) and Christof Weiß (christof.weiss@audiolabs-erlangen.de), 2020/21

1. Specify parameters and paths

2. Read and process audio files

3. Validate the processing pipeline

3.1 Overlay Plot

Show an overlay plot of CQT and pitch annotations

3.2 Sonification

Sonify the processed data as a stereo file (Left channel: original audio, Right channel: pitch annotations sonified with sinusoidals).